Known for the stellar performance, extreme scalability, and the most intelligent automation at its core, Exadata is the world-class database infrastructure for enterprise workloads that demand cost-effective disk storage, blazing-fast-flash, and dynamic memory acceleration. Exadata complements them all with its smart tiers deeply integrated with the Oracle Database.
In this second part of our deep dive into Exadata X11M, we explore three storage server options and the intelligent system software that drives them to deliver record-breaking performance across OLTP, analytics, and mixed workloads.
High Capacity (HC) Storage Servers: Tiered Intelligence for Scalable Performance
The High-Capacity Storage Server is the most deployed option in Exadata environments, and for good reason. It offers a powerful balance of cost-effective disk storage with flash-speed acceleration.
Each HC server includes:
- Twelve 22 TB disk drives, offering 264 TB of raw capacity
- Four 6.8 TB high-performance flash drives, providing 27.2 TB of Exadata Smart Flash Cache
- 1.25 TB of DDR5-based Exadata RDMA Memory (XRMEM) positioned in front of flash to supercharge reads
The architecture intelligently layers memory, flash, and disk to accelerate performance. Frequently accessed data is automatically cached in the XRMEM Data Accelerator, followed by the Smart Flash Cache, while less critical data resides on disk. This adaptive tiering ensures that 95%–99% of read requests hit memory or flash, delivering performance that rivals all-flash arrays at a fraction of the cost.
Even write operations are enhanced using Exadata Write-Back Flash Cache, which handles up to 13 million 8K write IOPS per rack, eliminating disk I/O bottlenecks for large-scale OLTP workloads.
Administrators can rely on intelligent automation or take control by defining SQL-level caching preferences. Unlike generic storage systems, Exadata understands Oracle workloads and intelligently bypasses flash for non-beneficial I/Os (like large backup writes), keeping performance optimized at all times.
Extreme Flash (EF) Storage Servers
For databases requiring ultra-low latency and maximum throughput, the Extreme Flash (EF) Storage Server delivers an all-flash configuration built for performance-first workloads.
Each EF server includes:
- Four 30.72 TB capacity-optimized flash drives, totaling 122.88 TB of raw flash capacity
- Four 6.8 TB performance-optimized flash cards, adding another 27.2 TB for Smart Flash Cache
- 1.25 TB of XRMEM, providing lightning-fast memory-tier acceleration
The EF configuration is designed to handle the most demanding OLTP and analytic workloads, delivering not just high throughput, but consistent, low-latency access to data. The integration of capacity flash and performance flash provides a flexible and high-performing environment for real-time databases and mission-critical systems.
High Capacity-Z (HC-Z) Storage Servers
For smaller environments or medium-scale workloads, Oracle offers the HC-Z Storage Server—a compact and efficient entry into the Exadata ecosystem.
Each HC-Z server includes:
- Six 22 TB disk drives for a total of 132 TB raw capacity
- Two 6.8 TB NVMe flash cards, enabling 13.6 TB of Smart Flash Cache
- 576 GB of XRMEM, placed strategically in front of flash
This configuration brings the same Exadata architecture—tiered storage, caching intelligence, and memory acceleration to organizations with more modest performance or capacity needs. It’s a perfect fit for dev/test environments, departmental databases, or consolidated workloads that demand enterprise-grade performance and manageability without the full-scale investment.
Accelerating Every Layer with Exadata System Software
Storage hardware is just one piece of the puzzle. What sets Exadata apart is its smart system software, purpose-built to offload database tasks and eliminate performance bottlenecks.
Here are key innovations that supercharge database processing:
Smart Scan Offloading
Instead of sending raw data to the database server, Smart Scan pushes SQL processing down to the storage layer. As a result:
- Only relevant rows and columns are sent back
- Scans are parallelized across all storage nodes
- Queries return faster, with less load on the database tier
Storage Indexes
Exadata maintains in-memory indexes inside the storage servers that track column value ranges. Before scanning disk, Exadata checks these indexes to skip irrelevant I/Os, massively reducing unnecessary disk reads.
- Fully automatic and persistent across restarts
- Requires no manual configuration
- Speeds up WHERE clause filtering dramatically
Smart Flash Log Write-Back + Pipelined Log Writes
OLTP workloads are highly sensitive to log write latency. Exadata introduces:
- Smart Flash Log Write-Back to route redo logs directly to flash, bypassing slow disks
- Pipelined Log Writes that exploit the RoCE network and RDMA to parallelize log writes and minimize latency
Together, they increase log write throughput up to 2.5x, with consistent low-latency commits.
Exafusion and Undo Block Sharing via RDMA
Exadata’s Exafusion Direct-to-Wire Protocol lets Oracle RAC nodes communicate over RDMA—bypassing OS and kernel overhead.
Exadata also reads remote undo blocks over RDMA to satisfy read consistency in OLTP, reducing inter-node latency and boosting RAC scalability.
AI-Powered Automation
Oracle Database on Exadata uses machine learning to:
- Implement Automatic Indexing, optimizing queries without manual tuning
- Enable Real-Time Statistics that adapt query plans on the fly based on data changes
This dynamic intelligence ensures peak performance without constant DBA intervention.
Oracle Exadata X11M brings the smarter software and deeper database integration. Whether you’re deploying the HC, EF, or HC-Z storage option, you’re leveraging a system that continuously adapts to your workload, balances performance with cost, and delivers unmatched throughput at every layer.
Stay tuned for Part 3, where we explore Exadata cloud deployment options, consolidation best practices, and hybrid scalability across on-prem and OCI.
Talk to our experts:
